VMware Cloud Community
MarcusOsb
Contributor
Contributor

Disk/datastore IOPS

What is considered, normal, high, slow and problematic disk IOPS levels? eg in ms?

0 Kudos
3 Replies
JarryG
Expert
Expert

"IOPS" is sure not measured in "ms". It is simply a number (input/output operations per second). Value depends on storage type (local/network), medium (hdd/ssd), concurrency of access (dedicated/shared), sequential/random, transfer size, etc, etc.

_____________________________________________ If you found my answer useful please do *not* mark it as "correct" or "helpful". It is hard to pretend being noob with all those points! 😉
0 Kudos
vThinkBeyondVM
VMware Employee
VMware Employee

I assume: You are looking ideal IO latency value in ms.

Just for your info..VMware offers one cool feature SIOC (Storage IO control), which throttles the the LUN  when there is IO contention, SIOC determines IO contention based on SIOC IO threshold(i.e. IO latency) & if data-store crosses the IO latency, SIOC starts throttling the LUN (i.e. distribute the IOPS based on IOPS shares assigned to each VM:::From UI or API you can assign IOPS shares to particular VM which is residing on datastore where SIOC is enabled)

Best practices to set this latency for SIOC is as follows:

  • For FC storage the recommended latency threshold is  20 – 30 MS
  • For SAS storage the recommended latency threshold is  20 – 30 MS
  • For SATA storage the recommended latency threshold is 30 – 50 MS
  • For SSD storage the recommended latency threshold is 15 – 20 MS

Learn more about SIOC here: Storage IO Control, the movie - Yellow Bricks     &  Storage I/O Control, the basics

There is another cool feature VMware offers is SDRS which can balance the IO load across cluster of datastores. (Read more :An introduction to Storage DRS      &   :Death to false myths: Storage IO Control = SDRS IO load balancing 

IOPS capacity depends on type of storage you are using: ie.FC(VMFS), iSCSI, NFS, local disks, vSAN, SSDs etc. IOPS values will be given by the storage vendors the way we get bike mileage values from bike vendors.

Hope this helps


----------------------------------------------------------------
Thanks & Regards
Vikas, VCP70, MCTS on AD, SCJP6.0, VCF, vSphere with Tanzu specialist.
https://vThinkBeyondVM.com/about
-----------------------------------------------------------------
Disclaimer: Any views or opinions expressed here are strictly my own. I am solely responsible for all content published here. Content published here is not read, reviewed or approved in advance by VMware and does not necessarily represent or reflect the views or opinions of VMware.

0 Kudos
jrmunday
Commander
Commander

I like to see storage latency below 5ms ... which is a completely different metric to throughput (MB/s) or IOPS.

These three metrics together allow you to measure performance baselines so that you know when you have an issue.

See example of my testing here;

EMC Symmetrix VMAX 40K testing on vSphere 5.0

Cheers,

Jon

vExpert 2014 - 2022 | VCP6-DCV | http://www.jonmunday.net | @JonMunday77
0 Kudos